An in-depth research was conducted on the real-time scheduling issues involved in the joint operation of runoff cascade hydropower station under high-intensity peak-load regulation and frequency regulation requirements, and proposed an intelligent control method for real-time load regulation of runoff cascade hydropower stations. Based on the characteristics of real-time power generation scheduling of cascade hydropower stations, ensuring the effectiveness and flexibility of scheduling strategies, taking into account the constraints of reservoir operation, grid safety, unit operation, and the impact of different water heads on unit power generation efficiency, aiming to achieve stable water level control of runoff type hydropower stations and rapid response of cascade total load control. Real time load Control is divided into two categories: dispatching mode and non dispatching mode. In the dispatching mode, constructed water power determination, stable water level, less load regulation, load balance economic dispatching models, to predict in advance whether the day ahead load planning curve needs to be adjusted and automatically track the planning curve. In non dispatch mode, automatically match the abnormal water level control model based on PID regulation and the stable water level model that meets the requirements of efficient power generation. At the same time, proposed a water level control solution method based on successive approximation and multi-objective dynamic programming, as well as a mixed integer programming model for reducing the difficulty of solving real-time hydropower scheduling for load allocation. The application of the proposed model in the cascade hydropower stations in the Shaxi River Basin shows that it can achieve high-precision automatic load adjustment, effectively reduce the number of regulation times by 6%, and only retain one dispatcher per shift, greatly reducing the work intensity of operators on duty. This research effectively improves the centralized control capability and economic operation level of the watershed, and has rich theoretical and practical significance for promoting the intelligent and intelligent construction of real-time load scheduling for cascade power stations.